Andersen v. Stability AI Ltd.
3:23-cv-00201
| N.D. Cal. | Jun 27, 2025Background
- Plaintiffs are artists suing AI companies (including Midjourney) for copyright infringement and Lanham Act violations relating to the use of their artwork in AI training datasets.
- The dispute centers on whether Midjourney must produce all datasets used to train its AI models or just those sourced from LAION, a public dataset.
- Plaintiffs’ operative complaint alleges direct copyright infringement based solely on the inclusion of their works in the LAION datasets, along with Lanham Act claims related to alleged false endorsement and trade dress.
- Plaintiffs made a discovery request seeking all of Midjourney’s training datasets, arguing these were crucial to their claims.
- Midjourney objected, arguing that non-LAION datasets are irrelevant to the current claims and that production would be unduly burdensome.
- The court is deciding a discovery dispute over the proportionality and relevance of non-LAION training datasets.
Issues
| Issue | Plaintiff's Argument | Defendant's Argument | Held |
|---|---|---|---|
| Scope of Relevance of Non-LAION Datasets | All datasets are crucial; copying from any source supports their claim | Only LAION datasets are relevant as claims are limited to alleged LAION use | Not relevant—limited to LAION datasets |
| Fair Use & Proportionality | Need full corpus to rebut 'small portion' fair use defense | Overall size not germane; targeted interrogatory sufficient | Full corpus not needed for fair use defense |
| Relevance to False Endorsement Claim | Implication of artist association relates to AI training content | Content of training data irrelevant to whether false endorsement occurred | Training data irrelevant to claim |
| Trade Dress & Likelihood of Confusion | Need to compare Plaintiffs' works with all training data | Comparison requires only outputs, not all training data | Training data unnecessary for claim |
Key Cases Cited
- Art Attacks Ink, LLC v. MGA Ent. Inc., 581 F.3d 1138 (9th Cir. 2009) (articulating likelihood of confusion test under the Lanham Act)
- Hachette Book Grp., Inc. v. Internet Archive, 115 F.4th 163 (2d Cir. 2024) (amount of copyrighted material made available is key in fair use)